基于MSET的电站风机故障预警技术研究

Research on Early Fault Warning Technology of Fan in Power Plant based on MSET Method

为了对电站风机进行故障预警并判断故障类型,在改进过程记忆矩阵构造的基础上进行MSET(多元状态估计)模型预测。利用AHP(层次分析法)求得各监测参数对风机故障预警作用的权重系数,并应用到相似度函数中,得到风机的相似度序列。利用滑动窗口法确定预警阈值,当发出预警信号后,分析各监测参数实际值和估计值间的相对误差,以此追溯风机的可能故障点。利用某电厂引风机数据进行故障模拟仿真,结果表明该方法能及时捕捉到风机异常并在追溯可能故障点的同时对故障类型进行判断。

In order to realize the early fault warning and estimate the fault type of fans in power plant,MSET(multivariate state estimation technique) model prediction was employed based on the improvement of process memory matrix. The weight coefficient of each monitoring parameter for fan early fault warning was calculated by AHP(analytic hierarchy process) and the results were applied to compute similarity sequence of the fans based on similarity function. Sliding window method was used to determine the warning threshold. The possible fault point of fans was determined by analyzing the relative error between actual value and estimated value when early warning signal was received. The actual data of induced draft fan was used to simulate faults. The results showed that the method proposed in this paper can identify the fault in time and determine the possible fault point and judge the type of failure.